Using AI Tools as a SEBI RA: Disclosure RulesSEBI research analysts can use AI tools, but they are solely responsible for client data, research based on AI output and legal compliance. They must also disclose to clients how far they use AI. This guide covers Regulation 24(7) and 19(vii), the wider Regulation 16C, and how to disclose and keep records. ![]() Last updated: 21 October 2026 Yes, a SEBI research analyst can use AI tools for research, but the RA carries full responsibility for them. Regulation 24(7) of the SEBI (Research Analysts) Regulations, 2014, in force since 16 December 2024, makes an RA who uses AI tools, at any scale, solely responsible for the security, confidentiality and integrity of client data, the data used for research, research services based on AI output, and compliance with the law. Regulation 19(vii) requires you to tell clients how far you use AI in your research services. You make that disclosure with your terms and conditions, and again whenever your use changes. Running research, onboarding and client records in one CRM built for research analysts makes both duties easier to show. This guide explains those rules, the wider AI rule that applies to all SEBI-regulated entities, and how to disclose and document AI use in practice. It is a summary, not legal advice; check the current SEBI regulations and circulars. What do SEBI's rules say about AI tools for research analysts?Several provisions now apply together. The table sets out each one and what it requires. What does "solely responsible" mean in practice?It means you cannot shift blame to the tool or its vendor. If an AI output is wrong, or client data leaks through an AI service, SEBI will look to you. Regulation 16C says the same applies whether you built the tool or bought it from a third-party provider. Client dataDo not paste client names, PAN, KYC details or portfolios into public AI chatbots. Use tools whose data handling you have checked and documented. Research inputsYou are responsible for the data the AI uses. Check that prices, financials and news come from reliable, properly licensed sources. OutputsTreat AI output as a draft. A qualified person should check facts, figures and reasoning before anything goes to clients. Wider legal complianceYour responsibility covers "any law for the time being in force", which includes data protection and IT laws as well as SEBI rules. Vendor contractsWhere a vendor supplies the AI, record what data it receives, where it is stored, and how you can audit it. The vendor cannot take on your regulatory duty. How should an RA disclose AI use to clients?SEBI requires you to disclose the "extent" of AI use but does not prescribe wording. A useful disclosure tells clients which parts of your process use AI and where a person reviews the output. Include it in the terms and conditions clients consent to at onboarding, and update it when your use changes. The examples below are for illustration only. Have your compliance adviser approve your own text. Where to show itPut it in your terms and conditions, which clients must accept before you charge fees. Repeat it on your website and in research reports if AI plays a large part. When to update itPara 1.7(b) asks for additional disclosure whenever required, so tell clients when you add AI to a new stage, such as drafting rationale. Can an AI tool issue recommendations on its own?The regulations do not ban AI-generated ideas. They do put conditions on every recommendation. It must be corroborated by relevant data and analysis (Regulation 20(4)). SEBI's FAQs say you need an adequate documentary basis supported by research, and that a chart with a time stamp is not enough. The RA remains solely responsible for research based on AI output. In practice, an automated signal pushed straight to clients without review is hard to defend. Keep a person in the loop and keep the rationale. Qualifications still applyResearch must be provided by qualified, certified individuals. An AI tool does not replace that requirement. Marketing claimsThe advertisement code bars implying percentage accuracy or assured outcomes. "AI-powered 90% accurate calls" breaks it twice. See our guide to the SEBI advertisement code for RAs. What records should you keep when using AI?Regulation 25 requires the signed and dated research report, the recommendation and the rationale, kept for at least five years. When AI helps, it is sensible to also keep:
Beyond the Regulation 25 records, these are good practice, not listed SEBI requirements. They help you show your annual compliance auditor and RAASB that you met the "solely responsible" standard. Our guide to record keeping for research analysts covers the core list. Is more AI regulation coming?Probably. SEBI published a consultation paper on guidelines for responsible use of AI and machine learning in Indian securities markets in June 2025. It covered governance, testing, disclosures to investors and data security. Check sebi.gov.in for whether final guidelines have been issued and how they apply to RAs before you change your process. An AI-use checklist for research analysts
Use AI with a clear record behind itFinuro includes AI-assisted rationale for recommendations. It sits alongside the records you need: research report PDFs with report months, disclaimers on recommendation messages and reports, delivery logs, and an audit trail of who created or changed each item and when. Your analyst should still review every AI-assisted rationale before it goes out, and your AI disclosure stays your responsibility. See how Finuro's CRM for research analysts keeps research and client records together. Related Finuro features
Frequently asked questionsAnswers to common questions on this topic. Built for SEBI RAs and IAs Keep AI-assisted research on recordCreate recommendations with AI-assisted rationale, disclaimers, delivery logs and an audit trail in one place. ![]() |
Book a Demo
See how Finuro handles onboarding, recommendations, model portfolios, and billing for your advisory firm. Tell us a little about your practice and we'll walk you through the platform on a live call.


